Sedef Akinli Kocak, Ryerson University Andriy Miranskyy, IBM Canada Gülfem Işıklar Alptekin, Galatasaray University Ayşe Başar Bener, Ryerson University Enzo Cialini, IBM Canada ICT4S, Zurich, February 15, 2013 1
2 Agenda Objective Why this Study Matters Background Case Study Results and Discussions Conclusions Future Work
Objective To analyze the impact of transforming legacy systems with respect to new business demands on the energy consumption. To rise awareness of environmental sustainability among IT studies/ practices and green software. 3
Why this Study Matters Green IT is an ideal way to address the environmental issues in order to achieve environmental sustainability. The most significant and immediate reductions come from hardware solutions. Rare cases in which they focus on software. However, software also plays an important role having a considerable impact on energy consumption. Applying Green IT concept to the existing software to meet business demand or designing sustainable software is a complex task. Improving existing software functionality or leveraging legacy systems according to end-users requirements has a significant impact on environment. Trade-off between improving software functionality and reducing energy consumption of a software product needs to be investigated. 4
Background Effects of IT on environmental sustainability were structured mainly in three levels. Direct, indirect and rebound effect. Most of the studies almost completely ignore the rebound effect. Software development plays a specific role in creating rebound effects. A software system does not directly consume energy, it intensely affects the hardware functioning. Most of the studies have been done on the hardware and Life Cycle Assessment for computers, mobile phone networks, monitors, data center and power minimization by architectural designs. There is a lack of models and descriptions regarding environmental sustainability in the area of computer software. "GreenSoft" model for software products, sustainable metrics and criteria for software design and development 5
Background Maintenance, modernization and replacement and software reengineering and modernization focus on retargeting, revamping, use of commercial components, source code translation, code reduction, and functional transformation. NASCIO survey in 2008 major driving force of moving towards modernization of IT systems and applications is green IT initiatives with 28 %. Keeping the software on demand and with high quality levels in respect to end-users requirements create a conflict in terms of software energy consumption. Each integrated quality feature is accompanied by increasing levels of energy consumption. 6
Case Study Description of Software under Study DB2 for Linux, UNIX and Windows Version 10.1 Focused on measuring efficiency of the DB2 workload in the presence and absence of the data compression feature. Data compression reduces storage requirements, improves I/O efficiency, and provides quicker access to the data from the disk. Adaptive data compression Feature Utilizes a number of compression techniques leading to significant reduction of storage space. Its usage can lead to CPU overhead associated with compression and decompression of the data. 7
Workload Description and Case Study Setup Reference workload TPC-H (Transaction Processing Performance Council) industry standard for measuring database performance. Workload: A set of business-oriented ad-hoc queries-1gb of raw data and 240 distinct queries. Lenovo ThinkPad T60 laptop with 3GB of RAM. Two hour interval. Counted the number of statements executed in a given time interval. Measured the amount of electricity consumed by DB2 running in the following two configurations: Without compression With compression The workload has been executed against each configuration three times to estimate measurement error. 8
Results and Discussions Table 1 : Workload and database characteristics with and without adaptive compression feature Database Feature Average Statement Count per hour Average Electricity Consumption (kwh) Space Consumption (Mb) Compression ratio Watts per statement per second No compression 415 ± 1.8% 0.035 ± 0.5% 1168.2 100% 302 Adaptive compression 814 ± 4.0% 0.045 ± 0.0% 455.6 61% 199 Compression ratio = (1-compressed size/uncompressed size) Work performance (in terms of energy) =Energy consumption/work completed 9
Results and Discussions Compression of data reduces tables size by 61% and leads to 97% performance improvement in comparison with the reference configuration. Slowest configuration of the database is with the No compression. Compression feature increases overall power consumption by 29% (in comparison with the reference configuration). Energy consumption per unit of work is reduced by 34%, due to increased query throughput. 10
Conclusions Considering environmental sustainability legacy system modernization regarding increasing the system functionality has a mixed effect on energy consumption. consumption per unit of time increasing; consumption per unit of work deceasing. The amount of savings for the production system will vary with system s setup and with associated workload. The need for energy consumption from software especially from legacy systems which has strong end-user demand and there is a strong need for more input from software engineering on how to greening software toward creating more sustaining environment. 11
Future Work Individual effects and joint effects of different features on energy consumption will be examined. Trade-off models can be examined in software development regarding environmental sustainability. 12
13